Research on Image Retrievzl Based on Color Efficient Match Kernels

As is known to all, color features have excellent robustness for each kind of transformation of an image so that they play important roles in image retrieval. Whereas the traditional color features lack the spatial distribution information so that lower retrieval performance is got. Spatial distribution information of color features is achieved by image blocking firstly. Secondly, efficient match kernels are introduced to map the local color features into color efficient match kernels of low dimension, which have better discriminate ability than the traditional color features. Finally, retrieval experiments are executed based on the color efficient match kernels. Experiments show that the color efficient match kernels can improve the image retrieval performance effectively.